Log in / Register

Home arrow Computer Science arrow Social Informatics
< Prev   CONTENTS   Next >

3 Related Work

There are many examples that take advantage of the fine spatial granularity offered by geolocated social media data to study cities in greater detail. In popular culture, such datasets have been used to create casual visualisations [18] to engage the lay audience [19]. Several prominent examples include maps that show key paths in transport infrastructure [20], track the use of different languages in cities [21, 22] and reveal the distribution of urban wealth [23]. Previous work in literature have also made use of geolocated datasets for a multitude of purposes such as studying or developing technologies to support land use analysis, crisis management and mobility.

Land Use Analysis. Applications that use geolocated social media data for land use analysis are generally concerned with identifying the type of activities that are most common in specific urban areas. Frias-Martinez et al described a straightforward procedure that combines a space partitioning technique with human deduction to identify changes in land use over time [24]. Livehoods, a project by Cranshaw et al [12] addresses the same issue but adopts an automatic technique to draw alternative neighbourhood boundaries by clustering nearby locations with similar social activities. Their approach illustrates how the fine spatial resolution offered by geolocated tweets can be used to reveal social-spatial divisions in cities. Kling and Pozdnoukhov [13] developed a more sophisticated system that addresses the same issue. However, their work differs from the former in that they extract a chronologically ordered set of keywords to provide analysts with time stamped contextual information of activity on the ground.

Crisis Management Systems. Apart from land use analysis, geolocated social media data also serve as a source of information in crisis management systems. The task addressed by analyst in this domain involves extracting information to monitor situations and explain how they evolve. Studies such as De Longueville et al's analysis of a forest fire near Marseille [25], and Prasetyo et al's investigation of how a severe haze affected the residents of Singapore [26], act as some instances to characterise how such data can be used as a quantifiable source of information in times of crisis.

Mobility Analysis. While there exists a diverse range of work that made use of social media data for land use analysis and crisis monitoring, relatively little has been done to tap its potential for understanding human mobility. Traffic and navigation is one application area where such data have been exploited for mobility analysis. Wei et al describes an approach for constructing routes that navigate popular landmarks in cities [27]. Likewise, Pan et al address the problem of detecting and describing traffic anomalies by monitoring changes in mobility behaviour [28]. Pan's approach however, does not infer routing information from geolocated social media data but analyses it for contextual information that may be useful to describe events occurring on the ground. Character profiling is another application area where geolocated social media data has been applied. Fuchs et al presents an analytical approach to extract knowledge about personal behaviour from geolocated social media data by classifying profiles based on movement trajectories [29]. Andrienko et al addresses a similar challenge but classifies profiles based on venues instead [30]. A similarity between the existing works that make use of geolocated social media data to derive mobility information is that they focus on very precise patterns. Our work differs from existing applications in that we are more concerned with identifying general flow pathways between locations in a territory rather than the actual transit route or specific points of interest. In this respect, recent work by Gabrielli et al [31] addresses a similar topic as us yet their intent was to identify semantic rather than spatial patterns.

Visual Analytics. There are two broad approaches to conduct data analysis. Automatic algorithms can be used to address well-defined task with a known set of steps [32] while visual analysis is often required to support explorative task that require human deduction and reasoning. Visual analysis is not new to spatial planning as the discipline has a tradition of using maps for thinking and reasoning [33, 34]. There are several visualisation techniques that are relevant for urban flow analysis. Minard's map of Napoleon's Russian campaign [35] is one of the earliest attempts at visualising flows. The map depicts the size of the French army by the width of a band on the map, and depicts the change in its numbers in relation to air temperature throughout the duration of the campaign. Tobler [36] provides some early examples of computer generated flow maps. Flow maps are maps that show the movement of objects from one location to another [37]. The objects that are represented vary by theme. Flow maps rely on a node-link type representation where lines of different widths are used to represent the direction and quantity of objects being moved. An alternative to the node link representation is an origin destination map [38]. Origin destination maps comprises of a set of origin destination matrices arranged in geographic order. The map is interpreted by tracing a point of origin to a corresponding destination in one of the other matrices. While benchmarks [39] have shown that the matrix representation outperforms the node-link representation in task such as search and quantity estimation, node-link representations are reported to be more effective at path finding, an important task in interpreting the direction and sequence of flows.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science